AI techniques for waste water treatment plant control case study: denitrification in a pilot-scale SBR

Authors:
Davide Sottara;Luca Luccarini;Paola Mello
Affiliations:
DEIS, Facoltà di Ingegneria, Bologna and ACS, PROT, IDR, Gestione Risorse Idriche, ENEA, Bologna;DEIS, Facoltà di Ingegneria, Bologna and ACS, PROT, IDR, Gestione Risorse Idriche, ENEA, Bologna;DEIS, Facoltà di Ingegneria, Bologna and ACS, PROT, IDR, Gestione Risorse Idriche, ENEA, Bologna
Venue:
KES'07/WIRN'07 Proceedings of the 11th international conference, KES 2007 and XVII Italian workshop on neural networks conference on Knowledge-based intelligent information and engineering systems: Part I
Year:
2007

Citing 6
Cited 1

A Predicate-Transition Net Model for Parallel Interpretation of Logic Programs

IEEE Transactions on Software Engineering
Fuzzy control as a fuzzy deduction system

Fuzzy Sets and Systems
Neural Networks: A Comprehensive Foundation

Neural Networks: A Comprehensive Foundation
Fuzzy inference to risk assessment on nuclear engineering systems

Applied Soft Computing
A hybrid neural network model for rule generation and its application to process fault detection and diagnosis

Engineering Applications of Artificial Intelligence
An intelligent remote monitoring system for artificial heart

IEEE Transactions on Information Technology in Biomedicine

A Pool of Experts to Evaluate the Evolution of Biological Processes in SBR Plants

HAIS '09 Proceedings of the 4th International Conference on Hybrid Artificial Intelligence Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

We propose to show how different AI techniques might be used in the development of a modular expert system, acting as a manager and advisor for the operation of a pilot-scale SBR urban wastewater treatment plant, fed with real sewage. The plant's depurative effectiveness and global biomass' health depend on the reactions of nitrification and denitrification, with the former taking place as soon as the latter is complete. Since the duration of the reaction cannot be predicted, we have trained an intelligent software to recognize the event analyzing the profiles of some available signals, namely pH, orp and dissolved oxygen, thus allowing us to optimize the process' yield and detect possible failures. Using a SOM neural network, the system has been trained to remember an adequate set of reference signals, which have been given meaning using Bayesian belief techniques. Eventually, using the formalism provided by logical languages, reasoning capabilities have been imparted to the system, allowing the real-time, online deduction of new pieces of needed information. Thanks to the integration of these techniques the system is able to assess the status of the plant and act according to the adequate known policies.